Why the air traffic control rebuild started with the wires

The week

On April 28, 2025, controllers at the Philadelphia facility that handles approaches to Newark, one of the busiest airports in the country, lost radar and radio contact with aircraft for somewhere between 30 and 90 seconds. Their screens went blank and their radios went silent. The gap sounds brief. In the sky above a crowded metropolitan area, it is a very long time. Controllers rely on radar to see where each aircraft is and on radio to tell pilots where to go. Without both, they are working blind and mute.

Newark endured weeks of delays afterward. The disruption was compounded by runway construction at the airport and by staffing shortages among controllers, so every equipment problem landed on a system already stretched thin.

On May 8, 2025, Transportation Secretary Sean Duffy announced a plan to build a new air traffic control system over several years. It would replace aging copper telecommunications lines, the decades-old wires that carry voice and data much like old landline telephones, with fiber, wireless and satellite links. Fiber-optic cable carries data as pulses of light and is faster and more reliable than copper. The plan would also install new radars and radios and upgrade facilities. Duffy asked Congress for funding.

Then, the day after the announcement, on May 9, a second brief outage hit the same facility.

Officials linked the failures to the telecommunications lines carrying radar data between New York and Philadelphia, some of them old copper. That detail is the heart of the story. The radar dishes and the controllers’ screens were not the weak point. The weak point was the connection between them: the wires that carry what the radar sees to the people who need to see it, many miles away.

For anyone responsible for a critical system, whether in aviation, emergency services or tax collection, the week offered an uncomfortable reminder. Systems rarely fail where the money and attention go. They fail in the plumbing in between.

Few government systems are as visibly critical as air traffic control, which made the May 2025 plan an unusually public case study in legacy modernization, the work of replacing old technology that an organization still depends on every day.

Agency chief information officers who run benefits systems, emergency dispatch, tax processing or court records face the same problem at smaller scale. Their systems mostly work, cannot be taken offline, and depend on parts nobody makes anymore. Modernizing them means rebuilding while the service keeps running.

Modernizing such a system has four distinct parts: the transport layer that connects its sites, the application functions that run on it, the cutover from old to new, and the people who operate and maintain both. The Newark episode exposed weakness in each, and the plan’s order of work started with the first.

What an air traffic control system is made of

Air traffic control is not one computer. It is a chain of equipment spread across many sites.

At one end are the sensors: radar dishes that bounce radio waves off aircraft to find their position, and radios that let controllers talk to pilots. At the other end are the people: controllers sitting at screens in facilities that may be many miles from the radar itself. In between is the transport layer, the telecommunications circuits that carry radar data and voice traffic from the sensors to the screens. A circuit, in this sense, is simply a dedicated connection leased from a telephone or network company.

On top of all that runs the software, the applications that turn raw radar signals into the dots and labels a controller sees, track flight plans and pass aircraft from one sector to the next.

The same anatomy applies to many public systems. A 911 center has call-handling equipment, mapping software, radio links to ambulances and police cars, and the circuits connecting them. A state unemployment system has online forms, eligibility rules, a payment engine and the networks linking offices and data centers. In every case, leaders tend to focus on the visible software and overlook the connections underneath.

The transport layer: where critical systems actually fail

The Newark failures did not start in a radar or a controller’s display. They started in the circuits carrying radar data and radio traffic between facilities. When those links dropped, controllers lost their picture of the sky.

On May 9, according to officials’ public accounts, the backup line did not take over as intended. That detail matters more than the age of the wire. A redundant circuit, a spare connection meant to take over automatically when the main one fails, that has never failed over under real conditions is an assumption, not a control.

Many agencies share this hidden dependency. Public safety answering points, the call centers that receive 911 calls, state data centers, motor vehicle offices and remote clinics often rely on carrier circuits that are now being retired as telecommunications companies withdraw copper services. These circuits rarely appear in modernization roadmaps. They belong to the network team or to the carrier, and they are paid for out of a telecom budget nobody reviews.

They should be inventoried first, with three facts for each: what it carries, whether a diverse path exists, and when failover was last tested end to end.

Diverse paths need checking, too. Two circuits bought from two carriers can still share a conduit, a central office or a bridge crossing. Two supposedly separate paths may in fact run through the same trench, and a single backhoe digging in the wrong place can cut both. Agencies should ask carriers for route diversity documentation and repeat the request after any network change.

A failover test that pulls the primary circuit during a staffed maintenance window, with operators watching, tells the agency more than any diagram.

The functions on top: replace by capability, behind owned interfaces

The plan covered radars, radios, automation software and facilities. Replacing all of that at once is rarely feasible. The workable approach is to define the functions the system must deliver, such as surveillance (knowing where every aircraft is), communications and flight data, and replace each behind stable interfaces.

An interface is the agreed way two parts of a system talk to each other: what information passes, in what format, and what each side promises to do with it. Interfaces work like electrical outlets: because plug and socket are standardized, you can replace a lamp without rewiring the house. When interfaces are well defined, one component can be swapped without disturbing the rest.

Agencies can apply the same pattern. An unemployment insurance system can be modernized by function: identity proofing (confirming applicants are who they say they are), claims intake, eligibility determination, payments. Each function can move on its own schedule if the interfaces between them are defined well.

The risk sits in those interfaces. Vague interfaces couple every function to every other, and the project slides back into a single large replacement, the kind of “big bang” project that has a long history of overrunning budgets and missing deadlines in government.

Interfaces also decide how dependent the agency becomes on one vendor. Documented, standard interfaces let a component be replaced later without reopening the whole system. Proprietary ones, private designs controlled by a single company, tend to make the first vendor the only realistic choice for every later phase. Agencies should own the interface specifications outright, write them into solicitations, and require vendors to build and test against them.

The cutover: two systems at once, funded explicitly

Air traffic control cannot pause for a migration weekend, and neither can 911 dispatch or a payments system. Modernization means running old and new in parallel, often for months or years, with a tested way back if the new component fails. That way back is called a rollback: the ability to switch to the old system quickly and safely.

Parallel running roughly doubles certain operating costs for a period and needs staff fluent in both systems. Budgets usually fund the new system and assume the old one simply stops. That assumption is where many projects go wrong.

Agency leaders should require that cutover costs appear as their own line items: duplicate operations, data reconciliation (checking that both systems produce the same answers), rollback rehearsal and the extended maintenance contract for the legacy platform. A plan without those lines is a plan to cut corners at the riskiest moment.

Program offices should also define cutover exit criteria in advance, such as a period of matched outputs between old and new, so the decision to switch off the legacy system is evidence-based rather than schedule-driven. Without agreed criteria, the pressure of a deadline, a budget year or a press announcement tends to win the argument.

The people: operators, maintainers and overseers

Newark was also a staffing story. Shortages among controllers and technicians magnified the effect of every equipment failure. New systems need trained operators, and old systems need maintainers until the day they retire. Both groups are scarce, and the maintainers are often close to retirement themselves.

A modernization plan therefore needs a workforce plan with named answers. Who keeps the legacy platform alive, and on what contract? Who learns the new one, and when? How does knowledge of undocumented interfaces move from the first group to the second? Losing the last person who understands an old interface mid-transition is a common and avoidable failure. In many agencies there is one engineer who knows why a particular old circuit is wired the way it is, and that knowledge lives nowhere but in their head.

The workforce plan also covers the agency’s own technical oversight. Large programs are usually delivered by contractors and integrators, the firms hired to coordinate many vendors and stitch their work together. The agency still has to judge their work, accept or reject deliverables and make architecture decisions. Agencies that outsource that judgment along with the delivery tend to discover problems late, when they cost most to fix.

A county dispatch center, wire by wire

Suppose a county’s 911 dispatch center runs on software installed many years ago. The system works, but the vendor has announced the end of support, and the local telephone company has warned that the copper circuits linking the dispatch center to the backup center across town will be retired.

The county’s first instinct is to buy a new dispatch application. A better plan follows the four parts.

Transport first. The county lists every circuit behind dispatch: the lines to radio towers, the link to the backup center, the connection to the state’s emergency network. For each, it records what it carries, whether a second path exists and when failover was last tested. It discovers that the “diverse” backup link runs through the same telephone exchange as the primary. It fixes that before touching the software.

Functions behind interfaces. The county separates call handling, mapping, unit tracking and records, and writes down how they exchange data. It puts those specifications into the procurement, so that a different vendor could replace the mapping tool later without starting over.

A funded cutover. The budget includes six months of running both systems, a reconciliation process comparing incident records, a rehearsed rollback and an extended support contract for the old platform. The switch-off happens only after an agreed period of matched results.

People. The county names the technician who maintains the old system, pairs him with the new system’s administrators and records what he knows about the old interfaces before he retires. It also keeps a small in-house team able to review the integrator’s work.

The work is unglamorous, and all of it reduces the chance that residents call 911 on the wrong day and reach a silent line.

Habits a non-technical leader can check

Agencies that handle legacy modernization well tend to share a recognizable set of habits. None of them depends on a particular technology or vendor, and all of them can be checked by a non-technical leader.

• The connections are on the map. There is a current inventory of every circuit behind critical services, with an owner, a carrier, the route it takes and the date of the last real failover test. The network is treated as part of the system, not as a utility bill.
• Backups are proven, not presumed. Failover is tested on a schedule, with the primary line actually cut while trained staff watch. Results, including failures, are reported to leadership.
• The roadmap is organized by function. Instead of one giant replacement date, there is a sequence of smaller moves, each delivering a working capability and each behind an interface the agency controls.
• The agency holds the blueprints. Interface specifications, data definitions and architecture documents belong to the agency and are written into every contract, so that later phases can be competed fairly.
• Transition costs are visible. Parallel running, reconciliation, rollback rehearsal and legacy support appear in the budget from the start, rather than being discovered when the money runs short.
• Switch-off is earned. The old system is retired only after it has been matched by the new one for an agreed period, and the decision is recorded with its evidence.
• Knowledge has a succession plan. Every legacy platform has a named maintainer, a named successor and a record of what only the maintainer knows.

These habits are not expensive in themselves. They mostly require someone with authority to insist on them, and to keep insisting when a deadline or a budget cycle makes shortcuts tempting. In practice, that someone is often an executive or a board-level committee, because the people closest to the work are the ones under the most pressure to cut corners.

Strong plans also avoid some familiar shapes: a single glossy launch date, a new system switched on over a weekend, or a modernization plan in which the network, the cutover and the workforce appear only as footnotes. Those plans may photograph well, but Newark is a reminder of where they tend to break.

The case for paying for the unglamorous parts

The case for this kind of discipline comes down to four points.

Risk. Critical systems fail most often in their connections and transitions, not in their headline components. Spending that ignores the wires and the cutover leaves the riskiest parts untouched.

Cost. Parallel running and rollback planning look expensive on paper. A failed cutover, emergency contracts or a public outage cost far more.

Independence. Owned interfaces keep future choices open and protect the agency from being locked into one supplier for decades.

Public trust. Newark showed how quickly a technical failure becomes a national story. Citizens do not distinguish between a radar fault and a telecom fault. They simply see that the system did not work.

Ignoring these points produces a familiar set of failures: treating telecom circuits as someone else’s budget, trusting backup lines that have never been tested, buying a monolithic replacement with proprietary interfaces, assuming the old system can be switched off on a fixed date, and letting the last expert retire before the knowledge is written down.

Where an agency’s own wires stand

1. Which carrier circuits sit behind our most critical services, and which of them run on copper that the carrier plans to retire?
2. For each critical circuit, is there a backup path that carriers have documented as not sharing a conduit, exchange or bridge with the primary, and when did we last cut the primary and watch the backup take over?
3. Are we replacing our system function by function, and who owns the interface specifications between those functions?
4. Does our budget include parallel running, reconciliation, rollback rehearsal and extended legacy support as separate lines, and what evidence must we see before we switch off the old system?
5. Who understands our legacy systems, how will we capture their knowledge before they leave, and do we have enough in-house expertise to judge our integrator’s work?

With hindsight

Money followed the plan. Congress provided a first tranche of $12.5 billion for air traffic control modernization in the budget reconciliation law signed on July 4, 2025, including $4.75 billion for telecommunications infrastructure and systems upgrades. That last figure stands out: a large share of the first payment went to exactly the layer where Newark had failed.

In December 2025, the Transportation Department selected Peraton to manage the overhaul, which it put at roughly $31.5 billion. Peraton’s role was that of an integrator, the single firm coordinating many parallel projects and vendors. Transportation Secretary Sean Duffy said about $20 billion more would be needed. At that point, he said more than a third of the old copper lines had been replaced with fiber or other modern connections.

By late September 2026, the department said most outdated copper in towers had been replaced with fiber, and it was asking Congress for a further $30 billion through a proposed Modern Skies Act. Those are the department’s own progress claims, and we would claim nothing more. Independent verification of how far the work has gone, and how reliable the new links are under stress, will matter more than announcements.

What can be said is that the approach the department took, starting with the network links and appointing a single integrator to manage many parallel projects, matches the structure described above. Wires first, then functions, with someone accountable for coordinating the whole.

For other agencies, the value of the story is that it is public. Many state and local governments face their own copper retirements, aging dispatch systems and thin technical benches. Agencies facing copper retirement in their own networks now have a public reference for why the wires come first. It is easier to make the case for an unglamorous circuit inventory or a funded parallel run when a budget committee has watched the same problem play out above one of the country’s busiest airports. The harder part, as ever, will be finishing the cutover and keeping the people who know how both systems work.